The multiscale defect detection for photovoltaic (PV) cell electroluminescence (EL) images is a challenging task, due to the feature vanishing as network deepens. To address this problem, an attention-based top-down and bottom-up architecture developed accomplish fusion. This architecture, called bidirectional attention pyramid (BAFPN), can make all layers of share similar semantic features. In BAFPN, cosine similarity employed measure importance each pixel in fused Furthermore, novel object...

The anomaly detection in photovoltaic (PV) cell electroluminescence (EL) image is of great significance for the vision-based fault diagnosis. Many researchers are committed to solving this problem, but a large-scale open-world dataset required validate their novel ideas. We build PV EL Anomaly Detection (PVEL-AD <sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">1, 2, 3</sup> ) polycrystalline solar cell, which contains 36 543 near-infrared images...

Image/video stitching is a technology for solving the field of view (FOV) limitation images/ videos. It stitches multiple overlapping images/videos to generate wide-FOV image/video, and has been used in various fields such as sports broadcasting, video surveillance, street view, entertainment. This survey reviews image/video algorithms, with particular focus on those developed recent years. Image first calculates corresponding relationships between images, deforms aligns matched then blends...

CSI-Based physical layer authentication is a promising candidate to achieve fast and lightweight for wireless communication. However, the current methods usually cannot initial are susceptible channel noise. Besides, learning-aided requires illegitimate state information (CSI) samples that difficult obtain. This letter proposes newly deep-CSI-based scheme solve above problems. We map CSI device's location further its authenticated identity via deep learning in static environment. Therefore,...

For the problem of action detection, most existing methods require that relevant portions interest in training videos have been manually annotated with bounding boxes. Some recent works tried to avoid tedious manual annotation , and proposed automatically identify videos. However, these only concerned identification either spatial or temporal domain, may get irrelevant contents from another domain. These are usually undesirable phase, which will lead a degradation detection performance. This...

Online inter-camera trajectory association is a promising topic in intelligent video surveillance, which concentrates on associating trajectories belong to the same individual across different cameras according time. It remains challenging due inconsistent appearance of person and lack spatio-temporal constraints between cameras. Besides, orientation variations partial occlusions significantly increase difficulty association. Targeting solve these problems, this work proposes an...

The login functionality, being the gateway to app usage, plays a critical role in both user experience and application security. As Android apps increasingly incorporate functionalities, they support variety of authentication methods with complicated processes, catering personalized experiences. However, complexities managing different operations processes make it difficult for developers handle them correctly. In this paper, we present first empirical study issues apps. We analyze 361 from...

Existing localization methods commonly employ vision to perceive scene and achieve in GNSS-denied areas, yet they often struggle environments with complex lighting conditions, dynamic objects or privacy-preserving areas. Humans possess the ability describe various scenes using natural language help others infer location by recognizing recalling rich semantic information these descriptions. Harnessing presents a potential solution for robust localization. Thus, this study introduces new task,...

Long non-coding RNAs (lncRNAs) are critical in the regulation of various biological processes. In recent years, plethora lncRNAs have been identified mammalian genomes through different approaches, and researchers constantly reporting regulatory roles these lncRNAs, which leads to complexity literature about particular lncRNAs. Therefore, for convenience researchers, we collected relationships built a database called 'LncReg'. This is developed by collecting 1081 validated lncRNA-associated...

Semantic segmentation is a challenging task that needs to handle large scale variations, deformations and different viewpoints. In this paper, we develop novel network named Gated Path Selection Network (GPSNet), which aims learn adaptive receptive fields. GPSNet, first design two-dimensional multi-scale - SuperNet, densely incorporates features from growing To dynamically select desirable semantic context, gate prediction module further introduced. contrast previous works focus on...

Context modeling or multi-level feature fusion methods have been proved to be effective in improving semantic segmentation performance. However, they are not specialized deal with the problems of pixel-context mismatch and spatial misalignment, high computational complexity hinders their widespread application real-time scenarios. In this work, we propose a lightweight Spatial Feature Calibration Network (CSFCN) address above issues pooling-based sampling-based attention mechanisms. CSFCN...

Vehicle re-identification (re-id) is a promising topic, which focuses on retrieving the same vehicles across different cameras. It challenging due to variations of illumination and camera viewpoints. To solve these problems, we present multi -attribute driven vehicle re-id approach learn discriminative representations. The proposed consists multi-branch architecture re-ranking strategy. extracts color, model, appearance features, explicitly leverages attribute cues enhance generalization...

Unmanned aerial vehicles (UAVs) can capture high-quality photos and have been widely used for large-scale urban 3D reconstruction. However, even with the help of commercial flight control software, it is still a challenging task non-professional users to full-coverage in complex environments, which normally leads incomplete In this paper, we propose novel path planning method reconstruction scenes. The proposed approach first captures photos, following an initial generate coarse model as...

Road damage detection (RDD) is indispensable for safe autonomous driving. Existing RDD models focus on designing feature representations following expert knowledge. However, collecting and labeling all types of samples time-consuming leads to insufficient training data. To alleviate the adverse effect few samples, a novel few-shot road detector (FSRDD) proposed in this paper detect rare damages. The FSRDD includes three stages. First, fully annotated abundant base classes are leveraged train...

Open-set object detection (OSOD) aims to detect the known categories and reject unknown objects in a dynamic world, which has achieved significant attention. However, previous approaches only consider this problem data-abundant conditions, while neglecting few-shot scenes. In paper, we seek solution for generalized open-set (G-FOOD), avoid detecting classes as with high confidence score maintaining performance of detection. The main challenge task is that few training samples induce model...

Two-color pyrometric methods have been widely used in noncontact temperature measurement area. However, it is difficult to get synchronous monochromatic images for two-color formula. Some researches use beam splitter obtain two or more optical paths capture the different images, but complex will bring spatiotemporal matching errors. Another method uses color camera Red, Green, Blue (RGB) channel as RGB substituting Dirac delta function spectral response result inaccuracy of results. In fact,...

Crowd counting is a challenging problem due to the diverse crowd distribution and background interference. In this paper, we propose new approach for head size estimation reduce impact of different scale noise. Different from just using local information distance between human heads, global people in whole image also under consideration. We obey order far- near-region (small large) spread size, ensure that propagation uninterrupted by inserting dummy points. The estimated further exploited,...

In recent years, the amount of remote sensing imagery data has increased exponentially. The ability to quickly and effectively find required images from massive archives is key organization, management, sharing image information. This paper proposes a high-resolution retrieval method with Gabor-CA-ResNet split-based deep feature transform network. main contributions include two points. (1) For complex texture, diverse scales, special viewing angles images, A network taking ResNet as backbone...

Due to occlusions and objects' non-rigid deformation in the scene, obtained motion trajectories from common trackers may contain a number of missing or mis-associated entries. To cluster such corrupted point based into multiple motions is still hard problem. In this paper, we present an approach that exploits temporal spatial characteristics tracked points facilitate segmentation incomplete trajectories, thereby obtain highly robust results against severe data noises. Our method first uses...